Data-Centric Trustworthy AI

Jan 13, 2025·
Dr. Seong Joon Oh

Abstract:

I will present recent research efforts and future directions from the Scalable Trustworthy AI group at the University of Tübingen. Our goal is to bridge the gap between academic research and practical applications to enable broader adoption of AI. Key topics include:

  • Out-of-distribution (OOD) generalisation
  • Explainability
  • Uncertainty estimation
  • Privacy and security I will highlight key contributions in each area. Additionally, I will discuss my personal interest in medical AI applications and introduce the concept of a “dream” AI assistant. This assistant aims to enhance the productivity of experts, such as doctors, while increasing the value provided to their clients, such as patients.

About Joon:

Seong Joon Oh is a professor at the University of Tübingen, where he leads the group on Scalable Trustworthy AI (STAI). In addition to his primary role, he serves as an advisor to Parameter Lab. His research interests focus on training reliable models, including explainable, robust, and probabilistic models, and developing cost-effective methods for obtaining necessary human supervision and guidance. Previously, Seong Joon Oh worked as a research scientist at NAVER AI Lab for 3.5 years. He earned his PhD in computer vision and machine learning in 2018 from the Max-Planck Institute for Informatics, under the supervision of Bernt Schiele and Mario Fritz. His doctoral research focused on the privacy and security implications of computer vision and machine learning. He holds a Master of Mathematics with Distinction (2014) and a Bachelor of Arts in Mathematics as a Wrangler (2013), both from the University of Cambridge.